FastSurfer代码复现

最新推荐文章于 2024-07-27 17:39:19 发布

jennieyy

最新推荐文章于 2024-07-27 17:39:19 发布

阅读量190

点赞数 1

分类专栏：调代码文章标签： python 深度学习图像处理

本文链接：https://blog.csdn.net/weixin_48111108/article/details/140610639

版权

调代码专栏收录该内容

2 篇文章 0 订阅

订阅专栏

FastSurfer官方发布的Tutorial代码是Colab的，本文将介绍如何在本地服务器运行FastSurfer快速分割

第一步：run_prediction.py

E:\Anaconda\envs\fastsurfer\python.exe D:\FastSurfer-dev\FastSurferCNN\run_prediction.py --t1 "'D:/FastSurfer-dev/Tutorial/140_orig.mgz'" 
One of the following three options has to be passed --in_dir, --csv_file or --t1 with an absolute file path. Please specify the data input directory, the subject list file or the full path to input volume
[INFO: run_prediction.py:  652]: Checking or downloading default checkpoints ...

Process finished with exit code 1

需要Edit Configuration设置运行参数

--t1 D:/FastSurfer-dev/dataset/140_orig.mgz --sd D:/FastSurfer-dev/output/

运行结果是会在output文件夹下生成一个文件夹140_orig.mgz下面会有

D:\FastSurfer-dev\output\140_orig.mgz\mri\aparc.DKTatlas+aseg.deep.mgz

第二步：新建一个python script，mgz转换成nii，并运行Tutorial里的plot_predictions

import os
import sys
from os.path import exists, join, basename, splitext
img = 'D:/FastSurfer-dev/dataset/vs_gk_10vs_gk_t1_refT1_x2_SR.mgz'
output = 'D:/FastSurfer-dev/output/vs_gk_10vs_gk_t1_refT1_x2_SR/mri/aparc.DKTatlas+aseg.deep.mgz'
import nibabel as nib
# conversion to nifti
data = nib.load('D:/FastSurfer-dev/output/vs_gk_10vs_gk_t1_refT1_x2_SR/mri/aparc.DKTatlas+aseg.deep.mgz')
img_nifti = nib.Nifti1Image(data.get_fdata(), data.affine, header=nib.Nifti1Header())
nib.nifti1.save(img_nifti, 'D:/FastSurfer-dev/output/vs_gk_10vs_gk_t1_refT1_x2_SR/mri/aparc.DKTatlas+aseg.deep.nii.gz')

data = nib.load("D:/FastSurfer-dev/dataset/vs_gk_10vs_gk_t1_refT1_x2_SR.mgz")
img_nifti = nib.Nifti1Image(data.get_fdata(), data.affine, header=nib.Nifti1Header())
nib.nifti1.save(img_nifti, "D:/FastSurfer-dev/dataset/vs_gk_10vs_gk_t1_refT1_x2_SR.mgz")

import nibabel as nib
import matplotlib.pyplot as plt
plt.style.use('seaborn-v0_8-whitegrid')
from skimage import color
import torch
import numpy as np
from torchvision import utils

def plot_predictions(image, pred):
    """
    Function to plot predictions from validation set.
    :param images_batch:
    :param labels_batch:
    :param batch_output:
    :param plt_title:
    :param file_save_name:
    :return:
    """

    f = plt.figure(figsize=(20, 20))
    n, h, w = image.shape

    mid_slice = n // 2
    image = torch.from_numpy(np.expand_dims(image[mid_slice+16:mid_slice+32, :, :], 1))
    grid = utils.make_grid(image, nrow=4)

    plt.subplot(311)
    plt.imshow(grid.numpy().transpose((1, 2, 0)))
    plt.title('Slices')

    grid2 = utils.make_grid(torch.from_numpy(np.expand_dims(pred[mid_slice+16:mid_slice+32, ...], 1)), nrow=4)[0]
    color_grid = color.label2rgb(grid2.numpy(), bg_label=0)
    plt.subplot(312)
    plt.imshow(color_grid)
    plt.title('Prediction')

    plt.subplot(313)
    plt.imshow(grid.numpy().transpose((1, 2, 0)))
    plt.imshow(color_grid, alpha=0.3)
    plt.title('Overlay_GT')

    for ax in f.axes:
      ax.grid(visible=None)
      ax.axis("off")
    plt.tight_layout()

    plt.show()

orig_data = nib.load(img).get_fdata() / 255
pred_data = nib.load('D:/FastSurfer-dev/output/vs_gk_10vs_gk_t1_refT1_x2_SR.mgz/mri/aparc.DKTatlas+aseg.deep.mgz').get_fdata()
plot_predictions(orig_data, pred_data)

查看运行结果